use smaller number of parallel senders
Created by: camdencheek
I think this is causing failures in CI because we can't send 1000 messages in 100 milliseconds on CI. I expect this is just because CI is more CPU-loaded than my local machine. I seem to remember other issues about atomics being slow on CI as well.
This just runs it with a smaller number of parallel workers so that it will hopefully fit in the batch time-window in a stable manner.